Evaluation of algorithms for interaction-sparse recommendations : neural networks don’t always win

Klingler, Yasamin; Lehmann, Claude; Monteiro, Joao Pedro; Saladin, Carlo; Bernstein, Abraham; Stockinger, Kurt

doi:10.48786/edbt.2022.42

Please use this identifier to cite or link to this item: https://doi.org/10.21256/zhaw-24616

Full metadata record

DC Field	Value	Language
dc.contributor.author	Klingler, Yasamin	-
dc.contributor.author	Lehmann, Claude	-
dc.contributor.author	Monteiro, Joao Pedro	-
dc.contributor.author	Saladin, Carlo	-
dc.contributor.author	Bernstein, Abraham	-
dc.contributor.author	Stockinger, Kurt	-
dc.date.accessioned	2022-03-17T09:33:20Z	-
dc.date.available	2022-03-17T09:33:20Z	-
dc.date.issued	2022-03	-
dc.identifier.isbn	978-3-89318-086-8	de_CH
dc.identifier.uri	https://digitalcollection.zhaw.ch/handle/11475/24616	-
dc.description.abstract	In recent years, top-K recommender systems with implicit feedback data gained interest in many real-world business scenarios. In particular, neural networks have shown promising results on these tasks. However, while traditional recommender systems are built on datasets with frequent user interactions, insurance recommenders often have access to a very limited amount of user interactions, as people only buy a few insurance products. In this paper, we shed new light on the problem of top-K recommendations for interaction-sparse recommender problems. In particular, we analyze six different recommender algorithms, namely a popularity-based baseline and compare it against two matrix factorization methods (SVD++, ALS), one neural network approach (JCA) and two combinations of neural network and factorization machine approaches (DeepFM, NeuFM). We evaluate these algorithms on six different interaction-sparse datasets and one dataset with a less sparse interaction pattern to elucidate the unique behavior of interaction-sparse datasets. In our experimental evaluation based on real-world insurance data, we demonstrate that DeepFM shows the best performance followed by JCA and SVD++, which indicates that neural network approaches are the dominant technologies. However, for the remaining five datasets we observe a different pattern. Overall, the matrix factorization method SVD++ is the winner. Surprisingly, the simple popularity-based approach comes out second followed by the neural network approach JCA. In summary, our experimental evaluation for interaction-sparse datasets demonstrates that in general matrix factorization methods outperform neural network approaches. As a consequence, traditional well-established methods should be part of the portfolio of algorithms to solve real-world interaction-sparse recommender problems.	de_CH
dc.language.iso	en	de_CH
dc.publisher	OpenProceedings	de_CH
dc.rights	http://creativecommons.org/licenses/by-nc-nd/4.0/	de_CH
dc.subject	Recommender system	de_CH
dc.subject	Machine learning	de_CH
dc.subject	Neural network	de_CH
dc.subject.ddc	006: Spezielle Computerverfahren	de_CH
dc.title	Evaluation of algorithms for interaction-sparse recommendations : neural networks don’t always win	de_CH
dc.type	Konferenz: Paper	de_CH
dcterms.type	Text	de_CH
zhaw.departement	School of Engineering	de_CH
zhaw.organisationalunit	Institut für Informatik (InIT)	de_CH
dc.identifier.doi	10.48786/edbt.2022.42	de_CH
dc.identifier.doi	10.21256/zhaw-24616	-
zhaw.conference.details	25th International Conference on Extending Database Technology, Edinburgh (online), 29 March - 1 April 2022	de_CH
zhaw.funding.eu	No	de_CH
zhaw.originated.zhaw	Yes	de_CH
zhaw.pages.end	486	de_CH
zhaw.pages.start	475	de_CH
zhaw.publication.status	acceptedVersion	de_CH
zhaw.publication.review	Peer review (Publikation)	de_CH
zhaw.title.proceedings	Proceedings of EDBT 2022	de_CH
zhaw.webfeed	Datalab	de_CH
zhaw.webfeed	Information Engineering	de_CH
zhaw.funding.zhaw	NQuest – Natural Language Query Exploration System	de_CH
zhaw.author.additional	No	de_CH
zhaw.display.portrait	Yes	de_CH
Appears in collections:	Publikationen School of Engineering

Files in This Item:

File	Description	Size	Format
2022_Klingler-etal_Recommender-System_EDBT.pdf	Accepted Version	440.43 kB	Adobe PDF	View/Open

Show simple item record

Klingler, Y., Lehmann, C., Monteiro, J. P., Saladin, C., Bernstein, A., & Stockinger, K. (2022). Evaluation of algorithms for interaction-sparse recommendations : neural networks don’t always win [Conference paper]. Proceedings of EDBT 2022, 475–486. https://doi.org/10.48786/edbt.2022.42

Klingler, Y. et al. (2022) ‘Evaluation of algorithms for interaction-sparse recommendations : neural networks don’t always win’, in Proceedings of EDBT 2022. OpenProceedings, pp. 475–486. Available at: https://doi.org/10.48786/edbt.2022.42.

Y. Klingler, C. Lehmann, J. P. Monteiro, C. Saladin, A. Bernstein, and K. Stockinger, “Evaluation of algorithms for interaction-sparse recommendations : neural networks don’t always win,” in Proceedings of EDBT 2022, Mar. 2022, pp. 475–486. doi: 10.48786/edbt.2022.42.

KLINGLER, Yasamin, Claude LEHMANN, Joao Pedro MONTEIRO, Carlo SALADIN, Abraham BERNSTEIN und Kurt STOCKINGER, 2022. Evaluation of algorithms for interaction-sparse recommendations : neural networks don’t always win. In: Proceedings of EDBT 2022. Conference paper. OpenProceedings. März 2022. S. 475–486. ISBN 978-3-89318-086-8

Klingler, Yasamin, Claude Lehmann, Joao Pedro Monteiro, Carlo Saladin, Abraham Bernstein, and Kurt Stockinger. 2022. “Evaluation of Algorithms for Interaction-Sparse Recommendations : Neural Networks Don’t Always Win.” Conference paper. In Proceedings of EDBT 2022, 475–86. OpenProceedings. https://doi.org/10.48786/edbt.2022.42.

Klingler, Yasamin, et al. “Evaluation of Algorithms for Interaction-Sparse Recommendations : Neural Networks Don’t Always Win.” Proceedings of EDBT 2022, OpenProceedings, 2022, pp. 475–86, https://doi.org/10.48786/edbt.2022.42.